Putting AQE Under the Microscope: Joe Burrow vs Jared Goff

Digging further into the specifics of adjusted quarterback efficiency (AQE) for a pair of counterintuitive results

Yesterday I released the 2.0 version of my adjusted quarterback efficiency (AQE) metric. AQE attempts to better isolate quarterbacks’ contributions to the value generated in the plays they’re involved in, rather than simply calculating the mean expected points added (EPA) on those plays.

I think the results largely aligned with what football observers believe: Patrick Mahomes is No. 1, Jimmy Garoppolo sees his numbers drop versus his top-five unadjusted efficiency, while Aaron Rodgers and Matthew Stafford are adjusted up for poor offensive surroundings.

One of the more perplexing adjustments was for Joe Burrow, an assumed elite, young quarterback getting better each season. Perhaps Burrow isn’t in the class of Mahomes or Josh Allen yet, but he isn’t perceived as being far off. Burrow’s AQE drops him to 14th among all quarterbacks, from eighth by the unadjusted numbers.

Even more head-scratching (and thanks to readers for pointing this out), is social-media punching-bag Jared Goff jumping Burrow according to AQE, though still falling to 12th from 10th versus unadjusted EPA per play.

AQE will never be right, but it will hopefully be useful. A lot of the negative perceptions that metrics cause could be assuaged (wishful thinking) if critics understanding them a little better (assuming they care to do so). Metrics are really good at capturing what they’re supposed to capture by their methodology, but less good when opinions and grand insights are applied to them, especially mixing with sometimes toxic fandom. Ranking 32 quarterbacks by any metric gives critics 496 unique opportunities to pair two quarterbacks who they believe to have poor relatives rankings and then discount the entire methodology.

Here I’m going to walk through the adjustments for Burrow and Goff, give context for what the final results say, and even reflect on what could be missing in AQE 2.0, i.e. adjustments for 3.0 and beyond.